Cyber attack detection with artificial intelligence techniques

Cyberattacks have become a significant security threat, affecting both individual users and corporate systems. The increasing complexity of attack types has rendered traditional security measures insufficient, thereby enhancing the role of artificial intelligence and optimization techniques in cybersecurity. In this thesis, feature selection—a crucial preprocessing step in data mining—is performed using novel heuristic algorithms such as the Zebra Optimization Algorithm (ZOA) on the CSE-CIC-IDS2018 dataset. Subsequently, performance analysis will be conducted using machine learning models, including k-Nearest Neighbors (kNN), Random Forest, Multi-Layer Perceptron (MLP), Long Short-Term Memory Networks (LSTM), and Recurrent Neural Networks (RNN). The impact of optimization-based feature selection on model performance will be examined in detail, and accuracy rates across different models will be compared. The findings will assess whether specific optimization techniques provide a significant advantage in attack detection. This study highlights the importance of optimization-based feature selection methods for early cyberattack detection and provides insights into the potential application of more advanced algorithms in cybersecurity.